1 Librerias y carga de datos

# Cargar librerías
library(tidyverse)
library(leaflet) 
library(leaflet.extras) # funcionalidades extra para el paquete leaflet
library(geojsonio) # leer archivo GeoJSON con los barrios de Barcelona
library(broom) # pasar formato polygon a tidy 
library(mapproj) # se necesita para visualizar mapas con ggplot
library(plotly)
library(viridis) # escala color
library(ggtext) # dar formato a títulos de los graficos
library(grid)
library(facetscales) # diferentes escalas en los facets
library(scales)
library(cartogram) # mapas raros
library(tmap)
library(sp)
library(sf) # convertir sp a sf. Se necesita para mapas raros con plotly
library(htmlwidgets) # quitar cursor de plotly

# Leer archivo
data <- read.csv("https://raw.githubusercontent.com/lau-cloud/BCN_OPEN_DATA/master/2019_accidents_gu_bcn.csv", encoding="UTF-8", stringsAsFactors=TRUE) %>%
  rename(Longitude=Longitud, Latitude=Latitud)
  
# Leer archivo GeoJSON
neighbourhoods <- geojson_read("https://raw.githubusercontent.com/lau-cloud/BCN_OPEN_DATA/master/neighbourhoods.geojson", what='sp')

2 Mapa con leaflet

# Map
leaflet() %>%
  
  addTiles() %>%
  # addTiles(group="OSM") %>%
  # addProviderTiles("Stamen.Watercolor") %>%
  # addProviderTiles("Esri", group="Esri") %>%

  # Add polygons
  addPolygons(data=neighbourhoods, color="black", weight=2.5, opacity=1) %>%

  # Extras
  # addSearchOSM() %>%
  # addReverseSearchOSM() %>%
  addResetMapButton() %>%

  # Add marker layer for each sector with corresponding group name
                  
  # Matí
  addCircleMarkers(data=data %>% filter(Descripcio_torn=="Matí"), color="green", 
                   group="Matí", radius=3, stroke=FALSE, fillOpacity=0.9,
                   popup=~paste0("<b>", Nom_districte,"</b>", "<br/>",
                                 "Mes: ", Nom_mes, "<br/>",
                                 "Dia de la setmana: ", Descripcio_dia_setmana, "<br/>",
                                 "Hora: ", Hora_dia)) %>%
                                    
  # Tarda                                                         
  addCircleMarkers(data=data %>% filter(Descripcio_torn=="Tarda"), color="orange", 
                   group="Tarda", radius=3, stroke=FALSE, fillOpacity=0.9,
                   popup=~paste0("<b>", Nom_districte,"</b>", "<br/>", 
                                 "Mes: ", Nom_mes, "<br/>",
                                 "Dia de la setmana: ", Descripcio_dia_setmana, "<br/>",
                                 "Hora: ", Hora_dia)) %>%
  
  # Nit                                                           
  addCircleMarkers(data=data %>% filter(Descripcio_torn=="Nit"), color="blue",
                   group="Nit", radius=3, stroke=FALSE, fillOpacity=0.9,
                   popup=~paste0("<b>", Nom_districte,"</b>", "<br/>", 
                                 "Mes: ", Nom_mes, "<br/>",
                                 "Dia de la setmana: ", Descripcio_dia_setmana, "<br/>",
                                 "Hora: ", Hora_dia)) %>%
  
  # Add layer controls for base and overlay groups
  addLayersControl(overlayGroups=c("Matí", "Tarda", "Nit")
                   # baseGroups=c("OSM", "Stamen.Watercolor", "Esri")
                   ) %>%
  
  # Add legend
  addLegend(position="bottomleft",
            colors=c("green", "orange", "blue"), 
            labels=c("Matí", "Tarda", "Nit"))

# Número de accidentes por barrio
n <-  data %>%
  count(Nom_barri) %>%
  rename(neighbourhood=Nom_barri)
  
# Introducimos el número de accidentes por barrio en el archivo GeoJSON
neighbourhoods@data <- plyr::join(neighbourhoods@data, n) 

l <- leaflet(neighbourhoods) %>% addTiles() 

pal <- colorNumeric("viridis", NULL)

l %>% addPolygons(color="grey", weight=1, fillColor=~pal(log10(neighbourhoods@data[["n"]])),
                  fillOpacity=0.5,
                  label=~paste0(neighbourhoods@data[["neighbourhood"]], ": ", neighbourhoods@data[["n"]]), highlightOptions=highlightOptions(weight=4)) %>%
  addResetMapButton() %>%
  addLegend(pal=pal, values=~log10(neighbourhoods@data[["n"]]), 
            opacity=0.5, title="Nº accidentes", position="bottomright",
            labFormat=labelFormat(transform=function(x) round(10^x)))

3 Heatmap con leaflet

# Map
leaflet() %>%
  
  addTiles() %>%
  # addTiles(group="OSM") %>%
  # addProviderTiles("Stamen.Watercolor") %>%
  # addProviderTiles("Esri", group="Esri") %>%
  # addProviderTiles("Wikimedia") %>%
  
  # Add polygons
  # addPolygons(data=neighbourhoods, color="black", weight=2.5, opacity=1) %>%

  # Extras
  # addSearchOSM() %>%
  # addReverseSearchOSM() %>%
  addResetMapButton() %>%

  # Add heatmap
  addHeatmap(data=data, lng=~Longitude, lat=~Latitude, radius=7)

4 Mapas choropleth

4.1 ggplot con labels

# Número de accidentes por barrio
accidentes <- data %>%
  group_by(NK_Any) %>%
  count(Nom_barri) %>%
  arrange(desc(n)) %>%
  rename(id=Nom_barri) 

# Pasamos el archivo geojson de formato polygon a tidy
tidygeojson <- tidy(neighbourhoods, region="neighbourhood")

# Labels en el centro de cada barrio
# https://stackoverflow.com/questions/9441436/ggplot-centered-names-on-a-map
labels <- aggregate(cbind(long, lat) ~ id, data=tidygeojson, FUN=function(x)mean(range(x)))

# Labels
ggplot() +
  geom_polygon(data=tidygeojson, aes(x=long, y=lat, group=id), fill="#69b3a2", color="white") +
  theme_void() +
  coord_map() +
  geom_text(data=labels, aes(long, lat, label=id), size=2)

4.2 Número de accidentes por barrio con ggplotly

# Project the map
projected_map <- spTransform(neighbourhoods, CRS("+init=epsg:3395"))

# Pasamos el archivo geojson de formato polygon a tidy
tidygeojson <- tidy(projected_map, region="neighbourhood")

# Unimos datos numéricos (accidentes por barrios) con datos geográficos (latitud y longitud)
mergeddata <- merge(tidygeojson, accidentes, by=c("id"), sort=TRUE)

# Labels del número de accidentes
# labels2 <- aggregate(cbind(long, lat) ~ n, data=mergeddata, FUN=function(x)mean(range(x)))

# Mapa con ggplot
map <- ggplot() +
  geom_polygon(data=mergeddata, aes(x=long, y=lat, group=id, fill=n, 
                              text=paste0(id,": ",n)),
               colour="black") +
  theme_void() +
  coord_map() +
  # geom_text(data=labels2, aes(long, lat, label=n)) +
  # geom_text(data=labels, aes(long, lat, label=id), size=2) +
  scale_fill_viridis_c(trans=scales::pseudo_log_trans(sigma=0.001)) +
  theme(legend.title=element_blank()) +
  labs(title="Accidentes de tráfico en Barcelona por barrio (2019)")

ax <- list(
  title="",
  zeroline=FALSE,
  showline=FALSE,
  showticklabels=FALSE,
  showgrid=FALSE
)

# Mapa con ggplotly
ggplotly(map, tooltip=c("text")) %>%
  config(displayModeBar=F) %>% 
  layout(xaxis=ax, yaxis=ax) %>% onRender("function(el, x) {Plotly.d3.select('.cursor-crosshair').style('cursor', 'default')}")

4.3 plotly

accidentes2 <- accidentes %>% rename(neighbourhood=id)
geojson <- rjson::fromJSON(file="https://raw.githubusercontent.com/lau-cloud/BCN_OPEN_DATA/master/neighbourhoods.geojson")

g <- list(fitbounds="locations", visible=FALSE)

fig <- plot_ly() %>%
  add_trace(type="choropleth",
            geojson=geojson,
            locations=accidentes2$neighbourhood,
            z=accidentes2$n,
            colorscale="Viridis",
            featureidkey="properties.neighbourhood")

fig <- fig %>% layout(geo=g) %>% 
  colorbar(title="Nº accidentes") %>% 
  layout(title="Accidentes de tráfico en Barcelona (2019)", xaxis=ax, yaxis=ax) %>%
  config(displayModeBar=F) 

fig

4.4 Facets

# Leemos archivo (2018)
data2 <- read.csv("C:/Users/xviva/OneDrive/Desktop/data viz bcn/accidentes2018.csv",
                 encoding="UTF-8", stringsAsFactors=TRUE) %>%
  rename(Longitude=Longitud, Latitude=Latitud) 
  
# Número de accidentes por barrio (2018)
accidentes2 <- data2 %>%
  group_by(Any) %>%
  count(Nom_barri) %>%
  arrange(desc(n)) %>%
  rename(id=Nom_barri,
         NK_Any=Any) 

# Unimos datos numéricos (accidentes por barrios) con datos geográficos (latitud y longitud)
mergeddata2 <- merge(tidygeojson, accidentes2, by="id", sort=TRUE)

# Datos de 2018 y 2019
data1819 <- rbind(mergeddata, mergeddata2)

# Mapa con ggplot
mapa <- ggplot() +
  geom_polygon(data=data1819, aes(x=long, y=lat, group=id, fill=n, 
                                  text=paste0(id,": ",n)),
               colour="black") +
  theme_void() +
  coord_map() +
  scale_fill_viridis_c(trans=scales::pseudo_log_trans(sigma=0.001)) +
  theme(legend.title=element_blank()) +
  labs(title="Accidentes de tráfico en Barcelona por barrio") +
  facet_wrap(vars(NK_Any))

# Mapa con ggplotly
ggplotly(mapa, tooltip=c("text")) %>%
  config(displayModeBar=F) %>% 
  layout(xaxis=ax, yaxis=ax)  %>% onRender("function(el, x) {Plotly.d3.select('.cursor-crosshair').style('cursor', 'default')}")

5 Cartogramas con ggplot

5.1 Continuous Area Cartogram

# Número de accidentes por barrio
n <-  data %>%
  count(Nom_barri) %>%
  rename(neighbourhood=Nom_barri)
  
# Introducimos el número de accidentes por barrio en el archivo GeoJSON
neighbourhoods@data <- plyr::join(neighbourhoods@data, n) 

# Project the map
projected_map <- spTransform(neighbourhoods, CRS("+init=epsg:3395"))

# Construct cartogram
cartogram <- cartogram_cont(projected_map, "n")

# Plot it 
  tm_shape(cartogram) + 
  tm_borders() +
  tm_polygons("n", style="jenks") +
  tm_layout(frame=FALSE, legend.position=c("left", "bottom"))

5.2 Non-contiguous Area Cartogram

# Construct cartogram 
cartogram <- cartogram_ncont(projected_map, "n")

# Plot it
tm_shape(projected_map) + 
  tm_borders() +
  tm_shape(cartogram) +
  tm_polygons("n", style="jenks") +
  tm_layout(frame=FALSE, legend.position=c("left", "bottom"))

5.3 Non-Overlapping Circles Cartogram

# Construct cartogram 
cartogram <- cartogram_dorling(projected_map, "n")

# Plot it
tm_shape(projected_map) + 
  tm_borders() +
  tm_shape(cartogram) + 
  tm_polygons("n", style="jenks", alpha=0.5) +
  tm_layout(frame=FALSE, legend.position=c("left", "bottom"))

5.4 Todo junto

# Continuous Area Cartogram
cartogram <- cartogram_cont(projected_map, "n")

map1 <- tm_shape(cartogram) +
  tm_borders() +
  tm_polygons("n", style="jenks", legend.show=FALSE) +
  tm_layout(frame=FALSE)

# Non-contiguous Area Cartogram 
cartogram <- cartogram_ncont(projected_map, "n")

map2 <- tm_shape(projected_map) + 
  tm_borders() +
  tm_shape(cartogram) +
  tm_polygons("n", style="jenks", legend.show=FALSE) +
  tm_layout(frame=FALSE)

# Non-Overlapping Circles Cartogram
cartogram <- cartogram_dorling(projected_map, "n")

map3 <- tm_shape(projected_map) + 
  tm_borders() +
  tm_shape(cartogram) + 
  tm_polygons("n", style="jenks", legend.show=FALSE, alpha=0.5) +
  tm_layout(frame=FALSE)

# Legend
legend <- tm_shape(projected_map) + tm_polygons("n", style="jenks") +
  tm_layout(frame=FALSE, legend.only=TRUE, legend.position=c("center", "center"))

tmap_arrange(map1, map2, map3, legend, nrow=1)

6 Cartogramas con plotly

6.1 Continuous Area Cartogram

# Continuous Area Cartogram
cartogram <- cartogram_cont(projected_map, "n")

plot_ly(span=I(1)) %>% 
  add_sf(
    data=st_as_sf(projected_map), 
    color=I("white"),
    hoverinfo="none"
  ) %>%
  add_sf(
    data=st_as_sf(cartogram), 
    stroke=I("black"),
    color=~n,
    split=~neighbourhood, 
    text=~paste(paste0(neighbourhood,": ",n)), 
    hoverinfo="text", 
    hoveron="fills"
  ) %>%
  layout(showlegend=FALSE) %>%
  config(displayModeBar=F) %>% 
  layout(xaxis=ax, yaxis=ax) %>% onRender("function(el, x) {Plotly.d3.select('.cursor-crosshair').style('cursor', 'default')}")

6.2 Non-contiguous Area Cartogram

# Non-contiguous Area Cartogram 
cartogram <- cartogram_ncont(projected_map, "n")

plot_ly(stroke=I("black"), span=I(1)) %>% 
  add_sf(
    data=st_as_sf(projected_map), 
    color=I("gray95"),
    hoverinfo="none"
  ) %>%
  add_sf(
    data=st_as_sf(cartogram), 
    color=~n,
    split=~neighbourhood, 
    text=~paste(paste0(neighbourhood,": ",n)), 
    hoverinfo="text", 
    hoveron="fills"
  ) %>%
  layout(showlegend=FALSE) %>% 
  config(displayModeBar=F) %>%
  layout(xaxis=ax, yaxis=ax) %>% onRender("function(el, x) {Plotly.d3.select('.cursor-crosshair').style('cursor', 'default')}")

6.3 Non-Overlapping Circles Cartogram

# Non-Overlapping Circles Cartogram
cartogram <- cartogram_dorling(projected_map, "n")

plot_ly(stroke=I("black"), span=I(1)) %>% 
  add_sf(
    data=st_as_sf(projected_map), 
    color=I("gray95"),
    hoverinfo="none"
  ) %>%
  add_sf(
    data=st_as_sf(cartogram), 
    color=~n,
    split=~neighbourhood, 
    text=~paste(paste0(neighbourhood,": ",n)),
    hoverinfo="text", 
    hoveron="fills"
  ) %>%
  layout(showlegend=FALSE) %>% 
  config(displayModeBar=F) %>%
  layout(xaxis=ax, yaxis=ax) %>% onRender("function(el, x) {Plotly.d3.select('.cursor-crosshair').style('cursor', 'default')}")

7 Waterfall

# Diferentes escalas para cada facet de la visualización
# scales_y <- list(
# mati=scale_y_continuous(limits=c(0, 2800), breaks=seq(0, 2800, 1000)),
# tarda=scale_y_continuous(limits=c(0, 700), breaks=seq(0, 700, 250)),
# nit=scale_y_continuous(limits=c(0, 700), breaks=seq(0, 700, 250)))

# Orden meses
data$Nom_mes <- factor(data$Nom_mes, levels=unique(data$Nom_mes))

# Matí
mati <- data %>%
  filter(Descripcio_torn=="Matí") %>%
  count(Nom_mes) %>%
  rename(start=n) %>%
  mutate(end=lead(start, n=1)) %>%
  mutate(Difference=end-start)

mati2 <- rbind(c("aa", 0, 338, 38), mati)
mati2$Nom_mes <- lead(mati2$Nom_mes, 1)
mati2$id <- seq_along(mati2$start)
mati2$type <- ifelse(mati2$Difference>0, "Más accidentes", "Menos accidentes")
mati2$Category <- rep("Mañana", 13)
mati2 <- mati2 %>% head(n=12)

# Tarda
tarda <- data %>%
  filter(Descripcio_torn=="Tarda") %>%
  count(Nom_mes) %>%
  rename(start=n) %>%
  mutate(end=lead(start, n=1)) %>%
  mutate(Difference=end-start)

tarda2 <- rbind(c("aa", 0, 410, 410), tarda)
tarda2$Nom_mes <- lead(tarda2$Nom_mes, 1)
tarda2$id <- seq_along(tarda2$start)
tarda2$type <- ifelse(tarda2$Difference>0, "Más accidentes", "Menos accidentes")
tarda2$Category <- rep("Tarde", 13)
tarda2 <- tarda2 %>% head(n=12)

# Nit
nit <- data %>%
  filter(Descripcio_torn=="Nit") %>%
  count(Nom_mes) %>%
  rename(start=n) %>%
  mutate(end=lead(start, n=1)) %>%
  mutate(Difference=end-start)

nit2 <- rbind(c("aa", 0, 97, 97), nit)
nit2$Nom_mes <- lead(nit2$Nom_mes, 1)
nit2$id <- seq_along(nit2$start)
nit2$type <- ifelse(nit2$Difference>0, "Más accidentes", "Menos accidentes")
nit2$Category <- rep("Noche", 13)
nit2 <- nit2 %>% head(n=12)


# Juntamos los tres data frame
data2 <- rbind(mati2, tarda2, nit2) %>%
   mutate_at(vars(start, end, Difference), as.numeric)

# Orden 
data2$Category <- factor(data2$Category, levels= c("Mañana", "Tarde", "Noche"))

# Visualización
ggplot(data=data2, aes(x=Nom_mes, fill=type)) + 
  geom_rect(aes(x=Nom_mes, xmin=id-0.45, xmax=id+0.45, ymin=end, ymax=start), alpha=3/5) +
  labs(y="Muertes", 
       title="<span style='color:firebrick;'><b> Incremento</span></b> y <span style='color:darkseagreen;'><b>disminución</span></b> de accidentes en 2019") +
  scale_fill_manual(values=c( "firebrick", "darkseagreen")) +
  geom_text(aes(x=id, label=Difference, y=end), position=position_stack(vjust=1), size=3) +
  theme_bw() +
  facet_grid(vars(Category), scales="free") +
  #facet_grid_sc(Category ~ ., scales=list(y=scales_y)) +
  theme(axis.text.x=element_text(angle=45, hjust=1),
        axis.title.x=element_blank(),
        legend.position="none",
        strip.text=element_text(colour='white', size=11),
        panel.grid.minor=element_blank(),
        plot.title = element_markdown(family = "sans",
                                  hjust = 0,
                                  vjust=3))